Stochastic model for outcome prediction in acute illness

The aims were to apply a stochastic model to predict outcome early in acute emergencies and to evaluate the effectiveness of various therapies in a consecutively monitored series of severely injured patients with noninvasive hemodynamic monitoring. The survival probabilities were calculated beginning shortly after admission to the emergency department (ED) and at subsequent intervals during their hospitalization. Cardiac function was evaluated by cardiac output (CI), heart rate (HR), and mean arterial blood pressure (MAP), pulmonary function by pulse oximetry (SapO2)(SapO2), and tissue perfusion function by transcutaneous oxygen indexed to FiO2,(PtcO2/FiO2), and carbon dioxide (PtcCO2)(PtcCO2) tension. The survival probability (SP) of survivors averaged 81.5±1.1%81.5±1.1% (SEM) and for nonsurvivors 57.7±2.3%57.7±2.3% (p<0.001p<0.001) in the first 24-hour period of resuscitation and subsequent management. The CI, SapO2,PtcO2/FiO2 and MAP were significantly higher in survivors than in nonsurvivors during the initial resuscitation, while HR and PtcCO2PtcCO2 tensions were higher in the nonsurvivors. Predictions made during the initial resuscitation period in the first 24-hours after admission were compared with the actual outcome at hospital discharge, which were usually several weeks later; misclassifications were 9.6% (16/167). The therapeutic decision support system objectively evaluated the responses of alternative therapies based on responses of patients with similar clinical-hemodynamic states.